Machine learning assisted quantum chemistry for Orquestra

Unsupervised machine learning has recently been introduced into the field of quantum many-body physics. A strategy based on generative models has been particularly successful in the data-driven learning of quantum states. In this proposal, we aim to adapt this technology to applications in quantum chemistry. The primary focus of this research will be on the reconstruction of molecular wavefunctions using data obtained from qubit-based quantum simulators, such as superconducting circuits or trapped ions. Such simulators have recently demonstrated the preparation of ground-state wavefunctions for simple molecules. Their measurement output can be used to train generative models, which have been shown to significantly facilitate the calculation of physical observables. Our strategy will begin by finding novel mappings from the fermionic Hamiltonians of the original molecules to qubit Hamiltonians amenable for reconstruction with two generative models:, the restricted Boltzmann machine (RBM) and the recurrent
neural network (RNN). Together with Professor Melko, the MITACS postdoc (Dmitri Iouchtchenko) will lead the research into these generative models, and develop the machine learning technology into a set of open source software libraries. In partnership with the team at Zapata led by Alejandro Perdomo-Ortiz, these libraries will be deployed as part of the Orquestra Platform.

Faculty Supervisor:

Roger Melko

Student:

Dmitri Iouchtchenko

Partner:

Zapata Computing

Discipline:

Physics / Astronomy

Sector:

Professional, scientific and technical services

University:

University of Waterloo